I’m reasonably convinced this is the best argument against LLMs. It’s the same reason Open is in OpenAI’s name. The understanding that centralizing the ownership of these tools is going to transform the world is widespread. That’s why the investment is so high. If power and wealth isn’t concentrated into these AI labs the investment isn’t worth it. Which means we have to ask ourselves if we want that. There’s plenty of futures which include LLMs and don’t include the centralization but they require a departure from our current trajectory. There was also no guarantee that programming and computing would become free like it is today.
> There's plenty of futures which include LLMs and don’t include the centralization but they require a departure from our current trajectory.
I don't think that's true at all. It's pretty clear that local models are the future of agentic coding, and everyone's been moving towards that goal.
It's also becoming clear that current models are much bigger than they really need to be. New research indicates that most transformer models can be shrunk significantly and still perform the same.
We definitely aren't there yet, but models that run on a single consumer GPU are getting better at a pretty fast pace. Model size keeps going down, efficiency keeps going up, and compute keeps getting faster and cheaper.
I really don't see a future where enormous datacenters are the only way to run a coding agent. Huge models might continue to be more performant, but the gap between that and a local model is closing quick.
Even back in the day you had to buy programming books and courses if you wanted to learn how to make the best code. That wasn't free. It's really not all that different from LLMs, you can code without them, but they're a good resource to help you when you're stuck. There's a billion free LLMs you can use, Grok, duck.ai, etc. you don't need money or a subscription to vibe code.
This is directly addressed by the author and part of the post? Tools were very expensive until gcc etc., and the internet made excellent free guides available.
most of the available services (anthropic, google, openai, xai, deepseek) have a free tier. you can't use it extensively, and have to wait... but its there.
programming has always gates... today is no different. arguably, there are quite a lot more free options than there were when i was coming up.
And prior to the desktop computer, you had to actually go work at a laboratory in order to do any programming whatsoever, which required significant amounts of educational and social access
What’s the point?
Writing deploying and delivering software has never been as accessible as it has ever been
Much like the author I learned on my own too and with a lot less help because I didn’t have a parent even guiding me through it
But that is not under threat and I’m not sure why people think it is
None of the arguments demonstrate even accidentally that there is LESS knowledge or fewer options.
This is the least locked in period and the better AI gets it will be an option to be even less locked in because you can just build and run everything yourself on your own hardware
Literally anyone can run the equivalent of an entire datacenter from 2000 on a handful of retired servers and old laptops at this point.
You can still write code without LLMs, much like you can write code without modern IDEs, or use C and assembly instead of higher-level languages. But there are significant differences between the skills you learn in the process, which I believe inhibits upward mobility.
There are those who started playing with computers when compilers were often more expensive than the computer they ran on, and those who came after you could download an entire "Unix" system and toolchain for free.
Entire industries and massive companies existed for tools and tooling that is now considered free and table-stakes. Heck, an operating system used to cost money and didn't come with much at all!
I was fortunate to grow up when the internet was full of free learning resources, but there was a time just before that when you really did need physical books to get beyond the basics.
I remember talking to people a couple decades older than me and being confused when they talked about having to buy compilers, too.
I don't know, there used to be IDE vendors that sold stuff to enterprises and offered freebies for educational purposes. Down the line there will be free offerings by the established players as well as OSS models you can run locally. Right now this is of course not enjoyable on existing hardware that a middle schooler might be using, put a bit more RAM into the MacBook Neo and this might change.
Programming is free if you do not consider price of your time. If you consider it, it is much higher than AI-associated costs. And even with AI-associated costs, it is still much cheaper than most other engineering professions, where physical realization is orders of magnitude more costly.
If you were a young person with plenty of time, the best way you could spend it would be learning to program without AI- whether you have money or not.
LLMs aren't a requirement though.. and if you're learning, you're probably better off without the things. I was pretty down and out after the .com bubble burst and was staying in a house a friend was renovating without internet access for a while... I learned C# from a big fat book and the beta command line compiler... for years, I knew the language and tools better than my peers.
You can't get that level of depth with an LLM... because you generally won't be digging in... for that matter, if you're vibe coding, you're even further removed from the details of how things are being done for better or worse.
LLM providers are interested in maximizing their profits, not minimizing your costs. The eventual goal of the providers, and the reason that they have trillion-dollar valuations, is because the objective is to capture the market and then increase the price to capture the value of any time you may be saving by using them. In other words, if your time savings amounts to $100 per hour by using LLMs, their goal is to eventually charge you $99.99 per hour for the privilege of using them.
An interesting side effect might be that only people locked out from using LLMs will learn how to program in the future, as vide coding doesn't teach you the fundamentals.
I know what you're thinking — when the calculator came about, being forced to compute in your head wasn't an advantage. But LLMs are different: a calculator is a strictly improved substitute for mental arithmetic, whereas an LLM is only an approximate solution — and it is far from clear whether LLMs will ever become a perfect solution, given the nuanced challenges around context management, interpreting intent, etc.
> An interesting side effect might be that only people locked out from using LLMs will learn how to program in the future, as vide coding doesn't teach you the fundamentals.
While thinking about/working with LLMs, I've been reminded more than once of Asimov's short story Profession (http://employees.oneonta.edu/blechmjb/JBpages/m360/Professio...). In it, no one goes to school: information is just dumped into your brain. You get an initial dump of the basics when you're a kid, and then later all the specialty information for your career (which is chosen for you, based on what your brain layout is most suited to).
The protagonist is one of a number of people who can't get the second dump; his brain just isn't wired right, so he's sent to a Home for the Feeble Minded to be with other people who have to learn the old-fashioned way.
Through various adventures he eventually realizes that everyone who was "taped" is incapable of learning new material at all. His Home for the Feeble Minded is in fact an Institute of Higher Studies, one of only a handful, which are responsible for all the invention and creation that sustains human progress.
> An interesting side effect might be that only people locked out from using LLMs will learn how to program in the future, as vide coding doesn't teach you the fundamentals.
This is the strange part for me. I'm one of those people that I assume are really common here on HN - I've been having fun coding on personal projects for a long time, somewhere circa 1978 iirc for me. Where I work we're starting to dip our toes into AI and vibecoding and I'm not a big fan. Even in my boring job the actual coding is the part I like the most. So I've taken a different tack. I've been prompting Claude to teach me how to do things, and that has worked out really well. Some basic info to start with, specific questions as needed, but I'm doing the work. I'm improving my productivity while still learning new things and having fun. Win-win for me.
Gemini has been teaching me embedded Linux, and last year ChatGPT taught me C#. All on the free tiers mind you. But I'm doing the work, it's just faster to ask questions than to dig through mailing lists and source code.
At work though, the pressure to move fast is too high, so I'm letting Claude Code so more work these days (nowhere near the majority, but I've found things i can trust it with).
I don't think i could deal with a paid plan myself given how unpredictable the models are and opaque the pricing is.
> when the calculator came about, being forced to compute in your head wasn't an advantage.
I'm not sure, whether that is true, because when educators want you to learn how to compute you are "locked out" of calculators. You don't get to use a calculator until after you learned basic arithmetic and you won't use a CAS when you are supposed to learn calculus.
Traveling used to be free. You could walk, run or swim anywhere you wanted. Now these cars and airplanes are ruining travel, they are expensive and hard to maintain. You have to buy tickets from vendors and the experience is completely different than walking.
One niggle... Basic wasn't really free either. At least not QB45, etc. It's wasn't super expensive, but it wasn't free... Also worth a mention is computers themselves back in the 80's and early 90's costs as much or more than the cars a lot of people were driving at the time. I remember seeing a used XT in 1993 for around $200 or so, which was cheap enough, but state of the art was an 80486 DX2 66...
Today, you can get an entry level sub for Claude Code or Codex for about $20/month... and while that may be really expensive in some parts of the world, it's not nearly as bad as a single state of the art compiler or dev tools in the early 90's over the course of a year or two until the next version came out. Let alone something like an MSDN subscription.
Tangentially related: the author of the blog is listening to LukHash. I remember the guys absolutely stunning cover of C-64 Bruce Lee theme: https://www.youtube.com/watch?v=iUHewyaavys
We programers have been depending on a centralized compute resources for much longer than LLMs.
For one, imagine having to discover StackExchange without Google search. Sure, those were gratis, but I'm not so sure programming was ever as free as the author says.
So far it still seems like it still is, but I think we will shortly have a lot of convoluted and very sparsely informational code that will be a PITA to read as a human.
I'm already reading a ton of LLM generated code by less skilled developers and understanding and reviewing it requires a paranoid attention to detail of the reader that I think you probably lack if these tools to generate large chunks of code seems like a good option to you at all.
Very tangential, but I could swear QBasic included an on-disk documentation system accessible from the editor. Maybe only later versions?
I also learned programming on QBASIC around the same time frame, but in my case it was mostly because all the old 90's computers were getting thrown away at that time, so there were plenty of parts around for a kid to learn about computers without breaking anything expensive.
It was pretty easy back then to find software that would work on those machines on the internet, too. I'm not so sure it would be as easy for young people to learn using yesterday's computers today.
I struggle to understand the "hackers" in HN vouching for proprietary LLMs. Like we have so much so good open source software that is top notch like linux, git, postgres, http, tcp/ip, and a long etc., and now we have these billionaires trying to make us use LLMs for coding at a hefty price.
I understand it from people like PG and the like, but real hackers? C'mon people
This depends on your definitions of "free" and "programming". Can you afford a PC? Can you afford internet to access documentation? A lot of people can't. Likewise, what is "programming" to you? Hello World in Python? Or fixing a driver bug in the Linux kernel? Those are worlds apart in terms of hardware requirements just to complete the build.
This is a big part of why I'm looking to develop a local LLM capability: having the hardware is a good start, but also developing the understanding on what the SoTA of local edge models can do, so we're not crippled if remote models stop being served, or at least some risk management.
It doesn't solve the problem of general LLM dependency (at the end of the day we gotta keep our brains sharp), but any LLM-based workflows aren't all of a sudden put at risk if we set up something that depends on it.
73 comments
> There's plenty of futures which include LLMs and don’t include the centralization but they require a departure from our current trajectory.
I don't think that's true at all. It's pretty clear that local models are the future of agentic coding, and everyone's been moving towards that goal.
It's also becoming clear that current models are much bigger than they really need to be. New research indicates that most transformer models can be shrunk significantly and still perform the same.
We definitely aren't there yet, but models that run on a single consumer GPU are getting better at a pretty fast pace. Model size keeps going down, efficiency keeps going up, and compute keeps getting faster and cheaper.
I really don't see a future where enormous datacenters are the only way to run a coding agent. Huge models might continue to be more performant, but the gap between that and a local model is closing quick.
Use an LLM with the equivalent knowledge of Linux kernel and tect editor? Or git clone them.
It's another state management scheme being sold to politicians and elder investors who don't know any better. Big tech 100% relies on elder abuse.
programming has always gates... today is no different. arguably, there are quite a lot more free options than there were when i was coming up.
What’s the point?
Writing deploying and delivering software has never been as accessible as it has ever been
Much like the author I learned on my own too and with a lot less help because I didn’t have a parent even guiding me through it
None of the arguments demonstrate even accidentally that there is LESS knowledge or fewer options.
This is the least locked in period and the better AI gets it will be an option to be even less locked in because you can just build and run everything yourself on your own hardware
Literally anyone can run the equivalent of an entire datacenter from 2000 on a handful of retired servers and old laptops at this point.
>
Even back in the day you had to buy programming books and courses if you wanted to learn how to make the best code. That wasn't free"Even before the extinction level meteor hit Ohio, there were tiny meteors hitting Earth all the time, it wasn't that safe either".
Entire industries and massive companies existed for tools and tooling that is now considered free and table-stakes. Heck, an operating system used to cost money and didn't come with much at all!
I remember talking to people a couple decades older than me and being confused when they talked about having to buy compilers, too.
This is exactly the kind of person that could be excluded by a programming culture that requires extensive use of LLMs.
You can't get that level of depth with an LLM... because you generally won't be digging in... for that matter, if you're vibe coding, you're even further removed from the details of how things are being done for better or worse.
I know what you're thinking — when the calculator came about, being forced to compute in your head wasn't an advantage. But LLMs are different: a calculator is a strictly improved substitute for mental arithmetic, whereas an LLM is only an approximate solution — and it is far from clear whether LLMs will ever become a perfect solution, given the nuanced challenges around context management, interpreting intent, etc.
> An interesting side effect might be that only people locked out from using LLMs will learn how to program in the future, as vide coding doesn't teach you the fundamentals.
While thinking about/working with LLMs, I've been reminded more than once of Asimov's short story Profession (http://employees.oneonta.edu/blechmjb/JBpages/m360/Professio...). In it, no one goes to school: information is just dumped into your brain. You get an initial dump of the basics when you're a kid, and then later all the specialty information for your career (which is chosen for you, based on what your brain layout is most suited to).
The protagonist is one of a number of people who can't get the second dump; his brain just isn't wired right, so he's sent to a Home for the Feeble Minded to be with other people who have to learn the old-fashioned way.
Through various adventures he eventually realizes that everyone who was "taped" is incapable of learning new material at all. His Home for the Feeble Minded is in fact an Institute of Higher Studies, one of only a handful, which are responsible for all the invention and creation that sustains human progress.
> An interesting side effect might be that only people locked out from using LLMs will learn how to program in the future, as vide coding doesn't teach you the fundamentals.
This is the strange part for me. I'm one of those people that I assume are really common here on HN - I've been having fun coding on personal projects for a long time, somewhere circa 1978 iirc for me. Where I work we're starting to dip our toes into AI and vibecoding and I'm not a big fan. Even in my boring job the actual coding is the part I like the most. So I've taken a different tack. I've been prompting Claude to teach me how to do things, and that has worked out really well. Some basic info to start with, specific questions as needed, but I'm doing the work. I'm improving my productivity while still learning new things and having fun. Win-win for me.
At work though, the pressure to move fast is too high, so I'm letting Claude Code so more work these days (nowhere near the majority, but I've found things i can trust it with).
I don't think i could deal with a paid plan myself given how unpredictable the models are and opaque the pricing is.
> when the calculator came about, being forced to compute in your head wasn't an advantage.
I'm not sure, whether that is true, because when educators want you to learn how to compute you are "locked out" of calculators. You don't get to use a calculator until after you learned basic arithmetic and you won't use a CAS when you are supposed to learn calculus.
Today, you can get an entry level sub for Claude Code or Codex for about $20/month... and while that may be really expensive in some parts of the world, it's not nearly as bad as a single state of the art compiler or dev tools in the early 90's over the course of a year or two until the next version came out. Let alone something like an MSDN subscription.
For one, imagine having to discover StackExchange without Google search. Sure, those were gratis, but I'm not so sure programming was ever as free as the author says.
https://idiallo.com/blog/programming-tools-are-free
I'm already reading a ton of LLM generated code by less skilled developers and understanding and reviewing it requires a paranoid attention to detail of the reader that I think you probably lack if these tools to generate large chunks of code seems like a good option to you at all.
Very tangential, but I could swear QBasic included an on-disk documentation system accessible from the editor. Maybe only later versions?
It was pretty easy back then to find software that would work on those machines on the internet, too. I'm not so sure it would be as easy for young people to learn using yesterday's computers today.
I understand it from people like PG and the like, but real hackers? C'mon people
Lots of people use locked down proprietary softwares and even GNU licenses have been criticized for being locked down
There are primitivist critiques of technology in general that show how technological systems require very restrictive global industrial systems
Pre-LLM eras had me hunting all over for poorly documented solutions to common problems, with vast amounts of limitations on what was possible
It doesn't solve the problem of general LLM dependency (at the end of the day we gotta keep our brains sharp), but any LLM-based workflows aren't all of a sudden put at risk if we set up something that depends on it.
You think it was always this easy to find high quality docs and packages written by others for free?